Apoe-ko mice (B6.129P2-Apoe^tm1Unc/J, Strain #002052) and their suggested wildtype controls (C57BL/6 J, Strain # 000664) at two different ages, 3 months and 9 months, were included (n = 16, 4 per group, all females). Additionally, Apoe-ko mice raised on a high-fat (HFD) diet for 24 weeks (Adjusted Calories Diet - 42% from fat, ENVIGO+++) starting at 4 months of age, and those on a regular diet (RD) were studied at 9 months old (n = 4–5 per group, all females). A total of 26 female mice were used in 6 groups: 3-month-old Apoe-ko, 9-month-old Apoe-ko, 3-month-old control, 9-month-old control, 9-month-old Apoe-ko with HFD diet, and 9-month-old Apoe-ko with RD, based on the resource equation method (Charan and Kantharia, 2013). Apoe-ko mice develop fatty streaks in the proximal aorta at 3 months of age. These lesions increase with age and progress to a more advanced stage, characterized by less lipid but more elongated cells, typical of pre-atherosclerotic lesions. In this study, different brain regions, eye tissues, and tear fluids were used to determine the expression levels of selected miRNAs at 3-month-old and 9-month-old ages. However, target mRNAs and protein markers were determined only in the neocortex-hippocampus and eye tissues.

This was done as previously described (Wijesinghe et al., 2023b). In summary, ketamine hydrochloride (80 mg/kg) and xylazine hydrochloride (10 mg/kg) were administered subcutaneously as anesthesia before tear fluid collection. Tear fluid was collected from both eyes using sterile Schirmer tear test strips and stored at −80°C. Thereafter, the mice were sacrificed, and the brain and eyes were promptly removed. The left hemisphere was dissected into five regions: neocortex with hippocampus (region 1), olfactory bulb (region 2), striatum-thalamus-hypothalamus (region 3), brainstem (region 4), and cerebellum (region 5). The left brain hemisphere and left eye were used for miRNA and total RNA extractions, while the right hemisphere and right eye were used for immunofluorescence staining.

The expression levels of 8 mature miRNAs including miRs -101a-3p, −125b-5p, −140-3p, −146a-5p, −15a-5p, −34a-5p, −342-3p, and -374c-5p were determined (Supplementary Table S1). The mature sequences of these miRNAs are similar between Mus musculus and Homo sapiens. Our previous work has already tested these miRNAs in the brain and eye samples of transgenic APP-PS1 AD mice (Wijesinghe et al., 2023b).

Target genes of these miRNAs were identified via TargetScan 7.2 (Agarwal et al., 2015) based on conserved seed regions. miRNA- target gene interactions are visualized using Cytoscape version 3.10.1 (Shannon et al., 2003). This includes amyloid beta precursor protein (App), presenilin 1 (Psen1), beta-site APP cleaving enzyme (Bace1), sortilin-related receptor LDLR class A repeats-containing (Sorl1), calcium voltage-gated channel subunit alpha1 C (Cacna1C), microtubule associated protein tau (Mapt), Rho associated coiled-coil containing protein kinase 1 (Rock1), glial fibrillary acidic protein (Gfap), signal transducer and activator of transcription 3 (Stat3), leukemia inhibitory factor (Lif), vascular endothelial growth factor A (Vegfa), autophagy related 12 (Atg12), sirtuin 1 (Sirt1), B cell leukemia/lymphoma 2 (Bcl2), brain derived neurotrophic factor (Bdnf), complement factor h (Cfh), and organic cation transporter novel type 1 (Slc22a4). Additionally, three target genes associated with Aβ clearance and inflammation, including aquaporin 4 (Aqp4), allograft inflammatory factor 1 (Aif1), and triggering receptor expressed on myeloid cells 2 (Trem2) were included. A total of 20 target mRNAs were screened in the neocortex-hippocampus and eye tissue (Supplementary Table S2).

Each miRNA tested in this study underwent functional enrichment pathway analysis, which included identifying target genes, the number of genes involved in each Reactome pathway, and significant p-values, based on strong experimental evidence using miRPathDB 2.0 (Kehl et al., 2020). The Database for Annotation, Visualization, and Integrated Discovery (DAVID) (Sherman et al., 2022) was used to determine the functional enrichment of target genes screened in this study, focusing on commonly involved Kyoto Encyclopedia of Genes and Genomes (KEGG) and Reactome pathways. Additionally, target genes of these 8 mature miRNAs, identified by miRTarBase (Huang et al., 2022) and validated through one of the three strong experimental methods (reporter assay, western blot, or qPCR) for Mus musculus, were subjected to Reactome pathway analysis.

The protocols were consistent with previous work (Wijesinghe et al., 2023b). miRNAs from tear fluids were extracted individually using the miRNeasy Serum/Plasma Kit, with 200 μL supernatant homogenized in QIAzol lysis reagent containing MS2 RNA. Cel-miR-39 RNA oligos were added before chloroform. miRNAs from brain and eye tissues were extracted using the miRNeasy Mini Kit, with eye tissues processed individually and brain tissues pooled by strain, age group and anatomical region. Cel-miR-39 was also added to tissue samples before chloroform. miRNA quantity and quality were assessed before cDNA preparation.

As previously published (Wijesinghe et al., 2023b), the assays involved multiple stages: poly(A) tailing, adaptor ligation, reverse transcription (RT), miR-Amp amplification, and real-time PCR. Each assay began with 2 μL of 10 ng miRNAs extracted from tissue or tear fluid. RT-qPCR was performed using the 7,500 Fast Real-Time PCR System (Applied Biosystems), with each sample and miRNA analyzed in a minimum of three replicates. Normalization was done using the spike-in control Cel-miR-39, as described in our previous study (Wijesinghe et al., 2023b).

Total RNA was extracted from pooled neocortex-hippocampus, and eye tissues (n = 4 for Apoe-ko and control groups, and n = 4–5 for diet-based groups) using the RNeasy^® Mini Kit. cDNA synthesis was performed with the SuperScript™ VILO™ cDNA Synthesis Kit. Approximately 10 ng of cDNA was used for each reaction. RT-qPCR was conducted on a 7,500 Fast Real-Time PCR System. Glyceraldehyde-3-phosphate dehydrogenase (Gapdh) was used as a reference gene for data normalization. Three sets of primer pairs were tested for each gene, and the most effective primer pair was selected for the experiment (Supplementary Table S2) (Untergasser et al., 2007). Each target gene was analyzed in at least three replicates per sample.

The right half of the brain and the right eye globe of the above animals were used for protein expression studies. Six-micrometer-thick mid-sagittal brain and eye cross sections were stained using a double immunofluorescence protocol as described in our previous work (Wijesinghe et al., 2023a). Both Apoe-ko mice (n = 4 in each age group: 3-month-old and 9-month-old) and control mice (n = 4 in each age group: 3-month-old and 9-month-old) were evaluated. For APP/Aβ peptide detection, an 88% formic acid pretreatment was applied for 5 min. This was followed by antigen retrieval, performed either with 0.05% proteinase K in Tris-EDTA buffer (pH 8.0) for 10 min at room temperature (RT) or by heat-induced antigen retrieval in citrate buffer (pH 6.0) for 10 min at a power level of 800 watts (Wijesinghe et al., 2023a). The screening utilized primary antibodies, including inflammatory markers such as glial fibrillary acidic protein (Gfap, rabbit polyclonal, Cat# Z0334), a marker for astroglia; ionized calcium-binding adapter molecule 1 (Iba1, rabbit polyclonal, Cat# 019–19,741), a marker for microglia/macrophage-specific calcium-binding protein; triggering receptor expressed on myeloid cells 2 (Trem2, rat monoclonal, Cat# MAB17291), a receptor found in microglia/macrophages; and 6E10 (mouse monoclonal, Cat# 803014, 1:200), which reacts to the 1–16 amino acid residues of both Aβ peptides and APP, with an epitope nearly identical in human and mouse species (Wijesinghe et al., 2023a). To validate the 6E10+ APP labeling, we used an additional knockout validated APP antibody (rabbit monoclonal, Cat# A17911) and the 12F4 antibody (mouse monoclonal, Cat# 805501) which is specific to Aβ 1–42 amino acid residues. Secondary antibodies used for fluorescence confocal microscopy were Alexa Fluor^® 488 goat anti-rabbit (Cat# 11070), Alexa Fluor™ 647 donkey anti-rabbit (Cat# 711605152), FITC goat anti-rat IgG2b (Cat# A110-111F), Alexa Fluor™ 546 goat anti-mouse IgG1 (Cat# 21123) and Alexa Fluor™ 546 goat anti-rabbit IgG1 (Cat# 11071). Negative control slides were processed without primary antibodies simultaneously.

Zeiss LSM800 confocal microscope equipped with ZEN 3.7 (blue edition) software was used for image acquisition.

In this study, we focused primarily on the neocortex-hippocampus, which is the most affected area in AD (Figure 2A; Supplementary Table S3). Eye tissue (Figure 2B; Supplementary Table S4) and tear fluid (Figure 2C; Supplementary Table S5) were also examined to explore the translational potential of retinal biomarkers, with an emphasis on tear-based biomarkers.

In the neocortex-hippocampus of 3-month-old Apoe-ko mice, all tested miRNAs were significantly downregulated compared to matched controls (Figure 2A; Supplementary Table S3). In the eye tissue, only two miRNAs showed significant dysregulation (−125b and −374c) between 3-month-old Apoe-ko mice and matched controls (Figure 2B and Supplementary Table S4). In the tear fluid, four miRNAs (−101a, −15a, −342, and −374c) were significantly downregulated in 3-month-old Apoe-ko mice compared to matched controls (Figure 2C and Supplementary Table S5), with the exception of miR-146a, which showed significant a 4-fold increase (p = 0.035). In 9-month-old Apoe-ko mice, five miRNAs (−125b, −146a, −15a, −342, and −374c) were significantly upregulated in both the neocortex-hippocampus and eye tissue compared to age-matched controls. Notably, two of these miRNAs (−146a and -15a) were similarly elevated in tear fluid. Additionally, miR-140 in both eye tissue and tear fluid, and miR-34a in tear fluid, were significantly upregulated in 9-month-old Apoe-ko mice compared to controls.

Over time, four miRNAs—125b, −146a, −342, and −374c—showed consistent and significant upregulation in the neocortex-hippocampus and eye tissue of Apoe-ko mice, with two of them (−125b, and −342) also being reflected in tear fluid. Additionally, miR-15a and miR-374c in neocortex-hippocampus, and miRNAs -101a, −140, −15a, and −34a in tear fluid showed, significant increases. In contrast, all tested miRNAs in the neocortex-hippocampus of control mice were significantly downregulated over time. However, in eye tissues, miR-374c showed significant upregulation, while the other miRNAs did not display any significant changes. Similarly, in tear fluid, miR-342 and miR-374c exhibited significant downregulation, while the other miRNAs remained unchanged.

Relative miRNA expression levels across different brain regions were compared between Apoe-ko mice and controls at 3 and 9 months of age (Figure 3A), as well as within the same group of animals over time (9-month-old vs. 3-month-old, Figure 3B). Overall, miRNA expression levels followed a consistent pattern across the five brain regions, with generally lower levels in 3-month-old Apoe-ko mice and higher levels in 9-month-old Apoe-ko mice.

After miRNA analysis, we determined target mRNAs associated with Aβ (App, Psen1, Bace1, and Sorl1), tau (Mapt, Rock1, and Cacna1C), inflammation (Gfap, Aif1, Trem2, Lif, and Stat3), Aβ clearance (Gfap, Aqp4, Aif1, and Trem2), apoptosis (Sirt1, Bcl2, and Atg12), and neuroprotection (Vegfa, Bdnf, Cfh, and Slc22a4) in the neocortex-hippocampus and eye tissue. Their selection was primarily based on both conserved seed regions (Figure 1C) and existing literature.

High-fat diet and ApoE deficiency have been studied in relation to retinal degenerative diseases. Because the retina is considered a surrogate tissue for studying AD, we investigated miRNA and mRNA levels in both the neocortex-hippocampus and eye tissue, as well as circulating miRNAs in tear samples, based on diet.

Our findings in mice on a high-fat diet indicated that in the neocortex-hippocampus (Figure 4A; Supplementary Table S3) and eye tissue (Figure 4B; Supplementary Table S4), the inflammatory miRNAs -125b, −146a, and −342 were significantly upregulated, while the anti-amyloidogenic/angiogenic miRNAs -101a, −15a, and −374c were significantly dysregulated, showing downregulation in the neocortex-hippocampus (−101a and −374c) and upregulation (−15a and −374c) in the eye tissues. However, none of the tested miRNAs showed significant differences in the tears (Figure 4C; Supplementary Table S5).

The majority of the target mRNAs tested showed no significant differences in the eye tissue of mice on a high-fat diet compared to those on a regular diet. However, Bace1, Gfap, and Cfh were significantly downregulated, while Bdnf was significantly upregulated in the high-fat diet mice (Table 3; Supplementary Figure S3). We were unable to determine all 20 target genes in the neocortex-hippocampus. Of the 15 mRNAs assessed, Bace1, Mapt, Gfap, Trem2, Atg12, Sirt1, Vegfa, Cfh, and Slc22a4 were significantly upregulated, while Rock1, Stat3, Lif, and Bdnf were significantly downregulated (Table 2).

Based on the differential expressions of target mRNAs, glial cells were notably altered in the Apoe-ko mice with ageing in both neocortex-hippocampus and eye tissue (Figures 3C,D). In addition to impaired lipid metabolism, the accumulation of APP/Aβ peptides could contribute to aberrant glial cell expression. Therefore, we examined glial cell proteins Gfap, Iba1, and Trem2, along with 6E10+ APP/Aβ peptides, in the neocortex-hippocampus and eye tissue.

Intraneuronal accumulations of 6E10+ APP/Aβ peptides were identified in the neocortex-hippocampus, with immunoreactivity varying across groups (Figure 5A; Supplementary Table S6). Brain endothelial cells at the meninges also stained positively for APP/Aβ peptides (red arrowheads, Figure 5A). 3-month-old Apoe-ko mice showed no significant difference compared to controls, but 9-month-old Apoe-ko mice had significantly higher levels (p = 0.023) (Figure 5D). Gfap immunoreactivity was significantly lower in 3-month-old Apoe-ko mice (p < 0.0001) than in controls, while no difference was seen in 9-month-old mice (Figure 5D). No age-related change in 6E10 immunoreactivity was observed in Apoe-ko mice, though controls showed a significant reduction (p < 0.0001). Conversely, Gfap immunoreactivity significantly changed with aging in Apoe-ko mice (p < 0.0001), with no such difference in controls. Spearman’s rank correlation revealed a significant negative association between 6E10 and Gfap immunoreactivities in 9-month-old Apoe-ko mice (p = 0.0185, r = −0.4585), but not in 3-month-old mice or controls (Figure 5E).

Supplementary Figures S4A–K illustrated differences in APP/Aβ peptide and astroglia expression in the neocortex-hippocampus of Apoe-ko and control mice at 3 and 9 months of age. Higher-magnification images revealed 6E10+ APP/Aβ peptides within neurons (Supplementary Figures S4B,F, indicated by white dashed box and white arrowheads) and endothelial cells at the meninges (Supplementary Figure S4H, indicated by red arrowheads), while white dashed circles marked 6E10-negative cells. The presence of 6E10+ APP in neurons (Supplementary Figures S4C,G, indicated by white dashed box and white arrowheads) and endothelial cells (Supplementary Figure S4I, indicated by red arrowheads) was confirmed using a rabbit monoclonal APP antibody. Additionally, the 12F4 antibody showed positive signals in brain endothelial cells at the meninges, but not within neurons (Supplementary Figures S4D,J, indicated by red arrowheads).

Iba1 expression and its colocalization with the Trem2 were observed in the neocortex-hippocampus (Figures 5B,C). Iba1+ microglia expressing Trem2 receptors were indicated using zoomed white dashed boxes and yellow arrowheads (Figure 5C). Both Iba1 and Trem2 immunoreactivities were significantly higher in 9-month-old Apoe-ko mice (p < 0.0001), with no significant difference in 3-month-old mice compared to age-matched controls (Figure 5F and Supplementary Table S6). Iba1 levels remained stable with aging in Apoe-ko mice but decreased significantly in controls (p < 0.0001). Conversely, Trem2 immunoreactivity increased significantly in Apoe-ko mice (p < 0.0001), remaining unchanged in controls. Spearman’s correlation showed a significant positive association between Iba1 and Trem2 in 3-month-old controls (p = 0.05, r = 0.3588), but not in Apoe-ko mice or 9-month-old controls (Figure 5G).

Supplementary Figures S5A–E was provided to show Trem2 expression with and without microglial colocalization in the neocortex-hippocampus of Apoe-ko and control mice at 3 and 9 months of age. Enlarged merged images (Supplementary Figures S5A–C) demonstrated the colocalization of microglia expressing Trem2 receptors (indicated by yellow arrowheads) and cells expressing only Trem2 (indicated by white dashed circles) in the dentate gyrus, CA1 region, and neocortex, respectively.

Overall, 6E10+ APP/Aβ peptides and the microglia/macrophage proteins Iba1 and Trem2 were significantly increased in the neocortex-hippocampus of 9-month-old Apoe-ko mice compared to 9-month-old controls. Within strains, Gfap and Trem2 levels showed an age-related increase in Apoe-ko mice, while 6E10+ APP/Aβ peptides and Iba1 levels exhibited a significant age-related reduction in controls.

Our immunolabeling data confirmed intraneuronal accumulation of 6E10+ APP/Aβ peptides within the retinal layers (Figure 6A). Notably, 9-month-old Apoe-ko mice displayed 6E10+ Aβ plaque-like deposition in the inner retina (Figure 6A, zoomed-in white circle). However, further validation with a larger number of samples and different antibodies for Aβ plaques is necessary. 3-month-old Apoe-ko mice exhibited significantly lower 6E10 immunoreactivity than controls (p = 0.018) (Figure 6C; Supplementary Table S7). In 9-month-old Apoe-ko mice, both 6E10 (p = 0.009) and Gfap (p = 0.0001) immunoreactivities were significantly higher compared to controls. Over time, 6E10 levels remained the same, but Gfap increased significantly in 9-month-old Apoe-ko mice (p = 0.026). In contrast, 6E10 and Gfap immunoreactivities were significantly reduced in 9-month-old controls (p < 0.0001 and p = 0.0008). Spearman’s correlation revealed a moderate positive association between 6E10 and Gfap immunoreactivities in 9-month-old Apoe-ko mice (p = 0.0003, r = 0.7030), with no significant correlations in 3-month-old Apoe-ko mice or controls (Figure 6D).

Supplementary Figures S6A–F was provided to show the differences in 6E10+ APP/Aβ peptide and astroglial expression in the retina of Apoe-ko and control mice at 3 and 9 months of age. Higher magnification images illustrated 6E10+ APP/Aβ peptide accumulation predominantly within the GCL (Supplementary Figure S6B). The presence of APP, predominantly within the GCL, was validated using a rabbit monoclonal APP antibody (Supplementary Figure S6C). The white dashed circles indicated a diffuse plaque-like depositions within the GCL in panels 6B and C. Additionally, 12F4 showed positive signals at the ILM but not within the neurons (Supplementary Figure S6D). The white dashed circle indicated a 12F4+ extracellular Aβ plaque-like deposition within the INL in panel D.

Iba1 and Trem2 colocalization were primarily observed in the NFL-GCL (Figure 6B, zoomed-in white dashed box). Iba1 and Trem2 immunoreactivities showed no significant differences in 3-month-old or 9-month-old Apoe-ko mice compared to age-matched controls (Figure 6E and Supplementary Table S7). Within Apoe-ko mice, both Iba1 and Trem2 levels remained stable over time (Figure 6E and Supplementary Table S7). In controls, Iba1 levels increased significantly (p < 0.0001), while Trem2 levels remained unchanged. Spearman’s correlation revealed positive associations between Iba1 and Trem2 immunoreactivities in 3-month-old Apoe-ko mice (p < 0.0001, r = 0.9184) and 3-month-old controls (p = 0.033, r = 0.4668). However, in 9-month-old Apoe-ko mice, a significant negative association was observed (p = 0.0103, r = −0.5468), with no significant correlation in 9-month-old controls (Figure 6F).

Supplementary Figures S7A,B illustrated Trem2 expression with and without microglia/macrophage colocalization within the retinal layers of Apoe-ko and control mice at 3 and 9 months of age. Trem2 and Iba1+ microglia/macrophage expression, with or without colocalization, was predominantly identified in the NFL-GCL.

Overall, 6E10+ APP/Aβ peptides and Gfap levels increased significantly in the retina of 9-month-old Apoe-ko mice compared to age-matched controls. In contrast, 6E10+ APP/Aβ peptides and Gfap levels were significantly reduced, while Iba1 expression was significantly increased, in 9-month-old controls compared to 3-month-old controls.